AMORE - Adaptive Model cOnfiguration fRamework for Software Effort Estimation Methods
DescriptionThe goal of the research is to enable software engineers to find the best prediction model configurations from past empirical data to accurately estimate software development efforts and costs required for their current software project, utilizing its successful model building experience. The proposal aims to study different prediction models with software project datasets to derive the best model configuration and develop criteria for an adaptive selection framework. The ultimate goal is to developing a novel AMORE framework to provide the best model configuration for the software effort estimation method to deal with the given datasets using transfer learning techniques, which addresses the above challenges in applying software effort estimation prediction models. The framework profiles different processing options for prediction model pre-processing, estimation optimization and post-processing steps needed to derive useful estimations. It systematically collects model configuration performance metadata, and exploits learned model combination and performance measures to adaptively formulate the best model configuration using Case-based Reasoning, and to select processing components to build and perform more accurate software effort estimation. It validates the model configurations to ensure that they are not subject to spurious effects, preventing the models incorrectly predict efforts required for the software project. The success of the AMORE project has the potential to standardize the development of high quality SEE models to ensure the importance of dataset handling in pre-processing options. This research project will develop a framework and implement a system to realize such a potential. Feasibility evaluation on model configurations are performed empirically in a more rigorous and scientific setting, provides more accurate software project effort estimates for practical use. The project will also support research students to study this important area in software engineering.
|Effective start/end date||1/09/15 → 22/06/18|